Bottom-up approaches to reading

One-Day Workshop on the Teaching of Reading – LTC Muhammadiyah University of Yogyakarta, 24 August 2002

3.2 The nature of the reading process

There are various views of the nature of the first language reading process, but, in general, they are within a scale with the two approaches called bottom-up approaches and top-down approaches as endpoints. Between these endpoints stands a set of approaches to which most researchers currently adhere Hudson, 1998, termed interactive approaches. The first group of approaches, the bottom- up approaches, are also termed outside-in models Cambourne, 1979 or data- driven models of reading Carrell Eisterhold, 1988; Cohen, Eysenck Le Voi, 1986; Silberstein, 1994, while the second group of approaches, the top-down approaches, are also called inside-out models Cambourne, 1979 or conceptually- driven models of reading Carrell Eisterhold, 1988; Cohen, Eysenck Le Voi, 1986; Silberstein, 1994.

3.2.1 Bottom-up approaches to reading

Hudson 1998:46 states that, according to bottom-up approaches, a reader constructs meaning from letters, words, phrases, clauses and sentences, sequentially processing the text into phonemic units that represent lexical meaning, and then building meaning in a linear manner. This approach assumes that the reading task can be understood by examining it as a series of stages that proceed in a fixed order from sensory input to comprehension. Rayner and Pollatsek 1989 assume that information is gained in a rather passive manner, processing is rapid and efficient, and the information that has been processed and stored in memory has little effect on how the processing occurs. According to the bottom-up approaches, the reader begins with the written text as the basis the bottom of the processing, and constructs meaning by absorbing and analysing small chunks of the text, gradually adding them to the next chunks until they become meaningful. In line with the bottom-up approaches, Taylor and Taylor 1983:116 point out that all reading processes must start with visual feature extraction of some kind. One-Day Workshop on the Teaching of Reading – LTC Muhammadiyah University of Yogyakarta, 24 August 2002 This process is identical to the visual feature extraction a human uses for perception in general. The eyes of a reader do not glide along the lines of print but instead perform a series of jumps called saccades or saccadic jumps with fixation pauses between them Taylor Taylor, 1983:52. It is further asserted Taylor Taylor, 1983:122 that as one reads, a target word is brought into the fovea the centre of the visual field in the retina by a saccadic jump. The eyes then fixate on the word for about a quarter of a second, during which time the image of the object is more or less stationary upon the retina. It is mainly during this fixation that a reader acquires information on the fixated word. At the end of the fixation, the eye saccades jumps to the next target word. Eye fixations tend to occur on informative words and clauses, and on the last words of sentences or paragraphs; regressions tend to occur on ambiguous or unexpected words. To find a word‟s meaning, skilled readers seem to use a visual route primarily and a phonetic route for special words, such as unfamiliar words. The visual route is a fast passive global process, whereas phonetic coding is a slow active analytic process. According to Taylor and Taylor 1983:232, “The visual path is a route to meaning, the phonetic path a route to remembering”. Taylor and Taylor 1983:277 point out that a reader processes a word as much as possible the moment it appears and then puts the results in working memory until enough words are accumulated to form a larger unit such as a phrase or a clause. Taylor and Taylor 1983:275 further suggest that in reading larger units of a text phrases, clauses, sentences, paragraphs, discourse, the ultimate goal of a reader is to comprehend its idea and to retain the idea as a gist. The reader must recognise most of the words in a clause, assign to them syntactic and case roles, find their meanings in the context, and organise them into larger processing units sentences, paragraphs, discourse. A clause or sentence is a major processing unit; at its end, the reader can usually integrate the information sufficiently to extract the gist. Soon after gist extraction, most individual words as well as the syntactic structure of a clause are purged from working memory to make space for new clauses Taylor Taylor, 1983:276. In doing this, the reader forgets the exact wording and retains the meaning Clark Clark, 1977:49. One-Day Workshop on the Teaching of Reading – LTC Muhammadiyah University of Yogyakarta, 24 August 2002 In summary, Taylor and Taylor 1983:116 point out that a reader of any script has the same goal to comprehend the content and retain the gist. To do so, the reader organises incoming material into larger units, distinguishes important from unimportant units, draws inferences to get the gist, and integrates gists to build higher-level gists to build the highest gist or the gist of the text. However, the “bottom-up” or “outside-in” processing theories have been criticised as being too simplistic and inflexible Byrnes, 1984:319; Stanovich, 1980:34; Hudson, 1998. In response, various “top-down” approaches to reading were developed. Some of these approaches are highlighted in the following sub-section. 3.2.2 Top-down approaches to reading According to Hudson 1998:47, the top-down approaches to reading assume that a reader approaches a text with conceptualisations above the textual level already in operation and then works down to the text itself. These approaches view the information processing circuit as being slower than assumed by the bottom-up approach as a result of memory capacity and mental limitations on the speed that information can be stored. Consequently, the reader makes continually changing hypotheses about the incoming information. The reader applies schemata, both formal schemata involving knowledge of rhetorical structures and conventions and content schemata involving knowledge of the world beyond texts to the text in order to create meaning that is personally and contextually sensible to the reader. Strong forms of these models assume that the reader is not text bound, but is sampling from the text in order to confirm prediction about the text message Smith, 1985; 1994. Goodman 1968 popularised this approach, calling reading a “psycholinguistic guessing game”. The key element was that reading was a psycholinguistic process that was an interaction between thought and language Goodman 1976. Goodman 1967 laid out the elements of language that he thought readers employed as they constructed meaning for the text they encountered. Goodman 1996:115 reiterates these elements when he says: One-Day Workshop on the Teaching of Reading – LTC Muhammadiyah University of Yogyakarta, 24 August 2002 By calling reading a psycholinguistic guessing game, I wanted to emphasize the active role of the reader in making sense of written language as a new key element in our understanding of the reading process. I wanted people to take distance from the view that reading is the accurate, sequential recognition of letters and words. I wanted them to understand that, in order to make sense construct meaning, readers: make continuous use of minimal information selected from a complex but incomplete and ambiguous text; draw on their knowledge of language and the world; use strategies of predicting and inferring where the text is going. In short, I wanted them to understand that readers engage in informed guessing as they read. Goodman‟s 1996 view of reading supports Smith‟s 1985:35 idea that reading depends more on what is behind the eyes - on non-visual information - than on the visual information in front of them. Summarising the gist of the top-down approaches to the reading process, Hudson 1998:47 suggests that readers use their knowledge of syntax and semantics to reduce their dependence on the print and phonics of the text. The reader makes guesses about the meaning of the text and samples the print to confirm or disconfirm. As Goodman, Watson and Burke 1996:9 claim, sampling, inferring, predicting, confirming, and integrating – always resulting in a personal construction of meaning – are the key operations or natural strategies within the reading process. According to Goodman, Watson and Burke 1996:5, as soon as a reader is confronted with print, heshe immediately starts sampling, inferring, and predicting. No reader uses all of the available cues. Heshe samples and infers the most significant cues and predicts what comes next. As a prediction is made, heshe tests it against hisher linguistic and conceptual knowledge to see if hisher prediction is meaningful. If the prediction matches the language and content of the print, the prediction is confirmed. If the prediction does not match the language and content of the print, the prediction is disconfirmed. If the latter happens, optional strategies are available to readers: Regress, reread, and pick up additional cues until the text makes sense. Stop, consider, and rethink why what is being read does not seem to make sense. Adjustments are made without rereading. Continue reading in order to build additional context; in so doing, generate enough understanding to decide why things do not make sense. Stop reading because the material is too difficult or not relevant Goodman, Watson Burke, 1996:7 One-Day Workshop on the Teaching of Reading – LTC Muhammadiyah University of Yogyakarta, 24 August 2002

3.2.3 Interactive approaches to reading